Extensions/ComfyUI-KieAPI
ComfyUI Extension

ComfyUI-KieAPI

ComfyUI custom node for Kie Nano Banana 2 image generation with optional reference images.

By Song367·Created 4 months ago·Updated 4 months ago· 0
Song367/ComfyUI-KieAPI
Nodes1
On cloudLocal install
Categoryapi/Kie
Stars0
Updated4 months ago
Readme

ComfyUI-KieAPI

Custom ComfyUI node plugin for Kie AI integrations.

Included in v1

  • Kie Nano Banana 2 node
  • text-to-image
  • optional reference image or reference image batch
  • API key via node input, KIE_API_KEY, or local config.json

Install

Place this folder under:

ComfyUI/custom_nodes/ComfyUI-KieAPI

Then restart ComfyUI.

Registry publishing

This repository is prepared for Comfy Registry publishing.

Required GitHub secret

Add a repository secret named REGISTRY_ACCESS_TOKEN with your Comfy Registry publisher API key.

Release flow

  • bump the version in pyproject.toml
  • push to main
  • manually run .github/workflows/publish_action.yml from GitHub Actions
  • the node is published through Comfy-Org/publish-node-action

Once published to the Registry, users can install it through ComfyUI-Manager without using Git URL installation.

Configure API key

The node resolves the Kie API key in this order:

  1. api_key node input
  2. KIE_API_KEY environment variable
  3. config.json in this plugin folder

To use a local config file, copy config.example.json to config.json and fill in your key:

{
  "api_key": "your-kie-api-key"
}

config.json is ignored by git.

Node inputs

  • prompt
  • api_key
  • reference_image (optional, accepts a single image or image batch)
  • aspect_ratio
  • resolution
  • output_format
  • google_search
  • poll_timeout_seconds

Notes

  • If you put api_key directly into the node, it may be stored in the workflow JSON.
  • Reference images are uploaded through Kie's base64 upload API before the generation task is created.
  • When reference_image receives a ComfyUI image batch, every image in the batch is uploaded to KIE in order.
  • The node polls Kie's common task detail endpoint until the task succeeds, fails, or times out.